KR20190053982A - Apparatus for analyzing customer emotion - Google Patents

Abstract

An emotion analysis device of the present invention includes: an analysis part identifying a dissatisfactory emotion of a customer by analyzing the voice of the customer; a classification part classifying a dissatisfaction level of the customer by using the dissatisfactory emotion; and a processing part executing an event set for the dissatisfaction level. Therefore, the present invention is capable of improving customer satisfaction while reducing dissatisfaction.

Description

{APPARATUS FOR ANALYZING CUSTOMER EMOTION}

The present invention relates to an apparatus for analyzing a customer's emotions.

Generally speaking, when a user accesses a service provider through a wired telephone or a mobile communication terminal, an interactive voice response (IVR) method accepts a combination of a user's voice input and a button that is selectively pressed, It provides a function to retrieve information and to automatically answer the retrieved information by voice, and to record voice for later processing. It provides various information such as customer center, micropayment system, opinion survey system, And so on. Such an IVR may be used in a similar manner to an ARS (Automatic Response Service).

Conventional IVR is a SI (System Integration) type business that has a large-sized equipment at a telephone office or an IDC (Internet Data Center) center and customizes scenarios and sound sources according to a specific business provider. There is a consistent problem with a rigid response that does not adapt adaptively.

In addition, IVR has a problem that it can not effectively resolve customer complaints due to limitations that it can not grasp the customer's feelings.

Therefore, it is necessary to grasp the customer's feelings in order to improve the satisfaction of the customer who is consulting through the call center, etc., and to provide a response scheme that is optimized for the identified feelings one step further.

Korean Patent Registration No. 0792990 discloses a technique of analyzing voice data of a receiver to determine the emotional state of a receiver, but it does not show a method of optimizing the emotional state of the receiver but optimizing the emotional state of the receiver.

Korean Patent Registration No. 0792990

The present invention is intended to provide an emotional analysis apparatus for customer response that grasps the customer's emotions and executes appropriate countermeasures in accordance with the emotions that have been grasped.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory and are not intended to limit the invention to the precise forms disclosed. Other objects, which will be apparent to those skilled in the art, It will be possible.

The emotional analysis apparatus of the present invention includes: an analysis unit for analyzing a customer voice to grasp a customer's feelings of dissatisfaction; A classifying unit classifying the complaint level of the customer using the complaint emotion; And a processor for executing an event set to the complaint level.

The emotion analyzing apparatus of the present invention classifies the emotion of the customer into classes and executes the events set for each class.

According to the present invention, by dividing the customer's emotion into a plurality of classes, an event in which the corresponding method according to the emotional state of the customer is specified can be appropriately executed.

Using an event that is properly executed according to the identified customer emotional state, the customer's dissatisfaction can be reduced and the customer's satisfaction can be improved.

Events that are run according to customer's level of dissatisfaction can be used to distinguish between a counselor's rating or a black consumer.

In addition, the effectiveness of the event set in each complaint level can be grasped by monitoring the change trend of the complaint grade changed due to the execution of the event.

According to the emotional analysis apparatus of the present invention, a customer's emotional analysis technique optimized for a response environment in which a consultant who is in consultation with a customer exists can be further provided.

1 is a block diagram showing an emotional analysis apparatus for a customer service according to the present invention.
2 is a schematic diagram showing the operation of the emotion analyzing apparatus of the present invention.
3 is a flowchart showing the emotional analysis method of the present invention.
Figures 4 and 5 are schematic diagrams illustrating call connection events.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. The sizes and shapes of the components shown in the drawings may be exaggerated for clarity and convenience. In addition, terms defined in consideration of the configuration and operation of the present invention may be changed according to the intention or custom of the user, the operator. Definitions of these terms should be based on the content of this specification.

1 is a block diagram showing an emotional analysis apparatus for a customer service according to the present invention.

The emotion analyzing apparatus shown in FIG. 1 may include an analyzing unit 110, a classifying unit 130, and a processing unit 150.

The analysis unit 110 can analyze the customer voice to grasp the customer's feelings of dissatisfaction. At this time, the analysis unit 110 can quantify the customer's feelings of dissatisfaction.

The classifying unit 130 can classify the customer's level of dissatisfaction using the complaint feeling expressed by the analyzing unit 110 and expressed numerically. For example, the classifier 130 may classify the digitized complaint feelings into one of a plurality of complaint classes. At this time, the event may be set according to the complaint level.

The processing unit 150 may execute an event set to the complaint level classified by the classification unit 130. [

The emotion analyzing apparatus of the present invention can be applied to a call center environment in which a customer and an agent talk. At this time, the analysis unit 110 may use the emotion grasping technique optimized for the call center environment.

For example, the analyzer 110 can accurately grasp a customer's feelings of dissatisfaction using waveform irregularities, waveform standard deviations, and detected negative words.

The analysis unit 110 can acquire a consultation voice of an agent who speaks with a customer. The analysis unit 110 can grasp the waveform irregularity (amplitude change, frequency change) using the result of the comparison between the waveform of the customer voice and the waveform of the consultation voice, and can quantify the complaint feeling.

Counselors related to counseling voice are users who have been educated on how to respond to customers, and can communicate with customers regularly and calmly, without emotionally responding to customer responses.

Therefore, the waveform of the consultation voice corresponding to the training of the counselor has a regular characteristic. On the other hand, the voice of the customer, which corresponds to the upbringing of the customer, is likely to be irregular depending on the emotional ups and downs of the customer. For example, the customer's voice becomes faster and harsher as the customer becomes excited, so that a change in the waveform's period shortening or a change in the amplitude of the waveform may appear. On the other hand, since the consultation voice will appear regularly and calmly despite the change of the customer voice, if the customer voice is viewed based on the consultation voice, the irregularity of the customer voice can be easily grasped.

The analyzer 110 can grasp the irregularity of the waveform of the customer voice based on the waveform of the consultation speech, and can quantify the irregularity.

For example, the analysis unit 110 may identify irregular intervals of the waveform of the customer voice with respect to the consultation voice, and may quantify the irregularities according to the number of the irregular intervals. For example, when one irregular interval occurs within a 5-second interval, the analysis unit 110 can display the customer's dissatisfaction feeling as 5. If two irregular intervals occur within the 5 second interval, the customer's dissatisfaction can be expressed as 10.

The complaint feeling quantified by the analysis unit 110 may be classified by the classifying unit 130 into one of a plurality of complaint grades.

The analysis unit 110 can grasp the standard deviation of the waveform and quantify the feeling of dissatisfaction using the comparison result between the standard waveform of the call voice and the customer voice obtained from a plurality of customers.

For example, the analysis unit 110 may define an average standard deviation by sampling 800 general customer voices. The analyzer 110 can give a high score to the customer's feelings of dissatisfaction as the difference between the standard waveform corresponding to the average standard deviation and the voice of the current customer is greater.

The analysis unit 110 can quantify the customer's feelings of dissatisfaction using the negative result of the negative speech included in the customer voice. Adjectives can include profanity, profanity, and so on.

The analysis unit 110 may classify adjectives such as profanity used by a person into a plurality of negative classes assigned different scores according to the severity. The analysis unit 110 can quantify the feeling of dissatisfaction using the score of the grade to which the negative word belongs.

For example, a first negative word corresponding to a mild profanity may be classified as a first negative grade with a score of 5. On the other hand, the second negative word corresponding to serious profanity can be classified as the second negative grade given 10 points.

The classifying unit 130 can quantify the customer's feelings of dissatisfaction using the score set in the negative grade of the negative speech included in the customer voice.

The analysis unit 110 may apply any one of analysis techniques such as waveform irregularity, waveform standard deviation, and detected negative words, or apply a plurality of analysis techniques simultaneously to quantify the customer's feelings of dissatisfaction .

When a plurality of analysis techniques are applied together, the analysis unit 110 may introduce a weight to accurately quantify the customer's feelings of dissatisfaction.

2 is a schematic diagram showing the operation of the emotion analyzing apparatus of the present invention.

The analysis unit 110 may include a plurality of analysis means for grasping the feeling of dissatisfaction and a collection means 115 for collecting a plurality of output values output from the plurality of analysis means.

At this time, the combining unit 115 assigns different weights to the output values of the analysis units, and can collectively grasp and quantify the customer's feelings of dissatisfaction using the output values of the weighted analyzing units.

For example, the analysis unit 110 may include a first analysis unit 111, a second analysis unit 112, a third analysis unit 113, and a collection unit 115.

The first analyzing means 111 can analyze the customer's feelings of dissatisfaction using the waveform irregularity. Specifically, the first analyzing means 111 can output a first output value obtained by digitizing the comparison result between the consultation voice of the consultant who speaks with the customer and the customer voice.

The second analyzing means 112 can analyze the customer's feelings of dissatisfaction using the standard deviation of the waveform. Specifically, the second analyzing means 112 can output a second output value obtained by digitizing the comparison result between the standard waveform of the call voice acquired from the plurality of customers and the customer voice.

The third analysis means 113 can analyze the customer's feelings of dissatisfaction using the negative language. Specifically, the third analyzing means 113 can output a third output value obtained by digitizing the negative speech included in the customer voice.

When the first output value of the combining unit 115 includes an adjective in a first section where the first output value satisfies the first set value or a negative section is included in a second section in which the second output value satisfies the second set value, 1 Weights can be added.

For example, if the waveform includes irregular intervals in the irregular interval, or if the irregular interval is included in the interval in which the standard deviation of the waveform is severe, it can be estimated that the customer is extremely excited. To reflect this fact, the combining means 115 may add a first weight to the third output value.

Also, the combining unit 115 may assign different weights to the first output value, the second output value, and the third output value. Specifically, the combining unit 115 may apply a second weight that is different from the third output value, the first output value, and the second output value to which the first weight is added.

For example, if negative words such as profanity are detected, the customer's complaint may be estimated to be serious. Therefore, a second weight (application rate) of 50% can be given to the third output value related to the negative speech.

In spite of the environment in which the consultation voice is quiet, if the irregularity of the customer voice continues or increases, it can be estimated that the excitement of the customer can not be calmly increased. Therefore, a second weight (application rate) of 40% can be given to the first output value related to the irregularity.

In the case of standard deviation, there may be a large margin of error depending on the volume and tone of each individual customer. Therefore, a second weight (application rate) of 10% can be given to the second output value related to the standard deviation.

The collecting means 115 may collect the first output value, the second output value, and the third output value to which the second weight is applied to quantify the customer's feelings of dissatisfaction.

The classification unit 130 may classify events corresponding to the customer's response plan into a plurality of levels and set a plurality of complaint levels so as to match events classified into a plurality of levels.

As an example, a first event may be provided to inform the customer that the complaint is increasing.

A second event may be arranged to guide the user to temporarily interrupt the conversation to alleviate the customer's dissatisfaction and calm the excitement.

A third event that provides a customer voice and a consultation voice to a second consultant other than the first consultant who is in consultation with the current customer may be provided.

A fourth event, which provides a second counselor's guidance to a first counselor who is in consultation with the current client, may be provided.

A fifth event may be provided to disconnect the call between the customer and the first consultant and to connect the call between the customer and the second consultant.

A sixth event may be provided that guides the customer to contact them after processing the request.

The seventh event that categorizes the customer as a black consumer can be provided.

At this time, the classifying unit 130 may classify the first event as the degree of interest set, the second grade of the event, the third grade of the event, the fourth grade of the excitement grade, The rage level for which the event is set, and the danger level for which the seventh event is set.

The classification unit 130 may classify the customer's dissatisfaction feeling into a degree of interest if the score output from the analysis unit 110, specifically, the combining unit 115 is 40 to 44 points.

If the classification unit 130 has a score of 45 to 49, the complaint feeling can be classified into the symptom grade.

If the classification unit 130 has a score of 50 to 64, it can classify the feeling of dissatisfaction into a warning grade.

If the classification unit 130 has a score of 65 to 69, it can classify the feeling of dissatisfaction into an excited grade.

If the classifying unit 130 has a score of 70 to 79, the complaint feeling can be classified as a complaint grade.

If the classification unit 130 has a score of 80 to 89, the complaint feeling can be classified into anger rating.

If the classification unit 130 is 90 points or more, the complaint feeling can be classified as a risk level.

If the current customer's complaint feelings are classified, the processing unit 150 can execute a complaint clearance event that lowers the customer's feelings of dissatisfaction from the current dissatisfaction level.

For example, the complaint resolution event may display a warning message on the counseling terminal 30 of the consultant who speaks with the customer if the complaint rating satisfies the set grade. This may be the first event described above.

Through the recognition of the warning message, the counselor can recognize that the current method of response is wrong and cope appropriately.

For example, a complaint resolution event may change the counselor to speak with the customer if the complaint rating meets the set rating. This may be the fifth event described above.

A customer who is not satisfied with the current counselor may resolve the complaint by talking to another counselor, for example, a senior manager.

The processing unit 150 can select a new consultant to consult with the customer according to the complaint level. The processing unit 150 may execute a connection event for making a call connection between the consulting terminal 30 of the selected consultant and the customer terminal 10 of the customer.

The processing unit 150 can monitor the change of the complaint level according to the execution of the complaint resolution event and store the monitoring result.

The processing unit 150 may generate a correction message for the complaint resolution event if the complaint level does not fall despite the execution of the complaint resolution event.

The complaint resolution event may be an event that satisfies the customer by lowering the customer's feeling of dissatisfaction. Therefore, customer complaints should be reduced through the execution of the complaint resolution event, and customer complaints may not be alleviated due to the erroneous complaint resolution event. In this case, the complaint resolution event needs to be corrected, and a correction message may be used to notify the counselor or manager of the fact.

The processing unit 150 may execute a consultation evaluation event for evaluating a consultant consulting with the customer using the complaint grade.

The processing unit 150 can evaluate each consultant considering the frequency with which the customer's complaint feelings occurred, the customer's feeling of dissatisfaction with the set grade, and the number of steps that were further increased. The manager can establish customer satisfaction measures such as retraining each counselor based on the evaluation results.

The processing unit 150 can execute a customer evaluation event for evaluating the customer using the complaint level. In spite of their kindness, they need to identify so-called black consumers who complain. The processor 150 can determine whether each customer corresponds to a black consumer on the basis of the frequency of complaints detected, the number of ratings of complaints raised, and the like.

The processing unit 150 stores the evaluation result of the customer evaluation event in each customer's personal information, and adds information indicating the black consumer to the personal information when the evaluation result satisfies the set value. The call center server can then call the second terminal of the black consumer counselor instead of connecting the customer terminal 10 to the first terminal of the general counselor when the call of the customer set as the black consumer is received.

3 is a flowchart showing the emotional analysis method of the present invention.

The emotion analysis method of the present invention may include a consultation and analysis step (S 510), a classification step (S 520), and an event execution step (S 530).

In the consultation and analysis step S 510, when a consultation is made between the customer terminal 10 and the consultation terminal 30 and between the customer voice and the consultation voice, a step of analyzing the customer voice to grasp the customer's feelings of dissatisfaction Lt; / RTI > The consultation and analysis step S 510 may be performed by the analysis unit 110.

The classifying step (S 520) may be a step of classifying the customer's dissatisfaction level using the grasped customer's dissatisfaction feelings. The classification step S 520 may be performed by the classification unit 130.

The event execution step (S 530) may execute various events such as a complaint resolution event, a counseling evaluation event, a customer evaluation event, a call connection event, etc. set to the complaint level. The event execution step (S 530) may be performed by the processing unit 150.

The consultation and analysis step S 510, the classification step s 520 and the event execution step S 530 may be repeated until the call with the customer is terminated (S 540).

Figures 4 and 5 are schematic diagrams illustrating call connection events.

When a call is made to the call center of the customer, a call is made to the first consultation terminal 31 of the currently available consultant.

If the customer's dissatisfaction is not detected, the processing unit 150 may finish the consultation call without executing the event.

If the customer's complaint rises to the level of interest during the consultation, the processing unit 150 may execute the first event informing that the customer's complaint is increasing. A warning message or the like may be displayed on the first consultation terminal 31 due to the execution of the first event.

The counselor can confirm the warning message displayed on the first consultation terminal 31 and take more care of consultation.

Despite the attention of the counselor, if the complaint class is not resolved and rises to symptom grade, the processing unit 150 may execute the second event.

Due to the execution of the second event, the first consultation terminal 31 may be prompted to stop consulting for a while.

After checking the message, the counselor can ask the customer for a patience and wait for the customer. The waiting time can be expected to reduce complaints as the customers organize their thoughts.

If the complaint level rises to the warning level despite the execution of the second event, it can be assumed that there is a problem with the consultation of the current counselor. To solve this problem, the processing unit 150 may execute the third event.

Due to the execution of the third event, the contents of consultation between the counselor and the customer can be provided to the different second counselor. The second counselor can monitor the present situation by monitoring the current consultation between the first counselor and the client. The second counselor may be an administrator-grade counselor managing multiple counselors.

If the level of complaints rises to the level of excitement despite the execution of the third event, a fourth event providing a guidance of the second counselor to the first counselor may be executed.

The first consultant can respond to the customer according to the guidelines of the second consultant displayed or aloud in the consultation terminal 30. [

If the complaint level rises to the complaint level despite the execution of the fourth event, the processing unit 150 can execute the fifth event.

According to the fifth event, the processing unit 150 disconnects the call connection between the first consulting terminal 31 of the first consulting client and the customer terminal 10 and disconnects the call connection between the customer terminal 10 and the second consulting company The call between the first and second consultation terminals 32 can be connected.

According to the execution of the fifth event, the customer who is talking to the existing first consultant can talk to the second consultant.

Despite the call with the second counselor, if the complaint level rises to anger rating, the current emotional state of the current client can be estimated to be unacceptable for any answer. At this time, the processing unit 150 can execute the sixth event.

Due to the sixth event, a guide message may be displayed to the second consultation terminal 32 that the customer will be contacted again after processing the request of the current customer.

Despite the execution of the sixth event, if the customer does not hang up and the customer's dissatisfaction level rises to a danger level, the seventh event that classifies the customer as a black consumer may be executed.

For example, according to the seventh event, the processing unit 150 can disconnect the call connection between the customer terminal 10 and the second consultation terminal 32 after transmitting the guidance message containing the content of the apology to the customer.

While the invention has been shown and described with reference to certain preferred embodiments thereof, it will be understood by those skilled in the art that various changes and modifications may be made without departing from the spirit and scope of the invention as defined by the appended claims. Accordingly, the true scope of the present invention should be determined by the following claims.

10 ... customer terminal 30 ... consulting terminal
31 ... first consultation terminal 32 ... second consultation terminal
110 ... analysis unit 111 ... first analysis means
112 ... second analysis means 113 ... third analysis means
115 ... collecting means 130 ... classifying section
150 ... processor

Claims (17)

An analysis unit for analyzing the customer voice to grasp the customer's feelings of dissatisfaction;
A classifying unit classifying the complaint level of the customer using the complaint emotion;
A processing unit for executing an event set to the complaint level;
And an emotion analyzing device.
The method according to claim 1,
The analyzing unit may quantify the complaint feeling,
Wherein the classifying section classifies the numerically expressed complaint emotion into one of a plurality of complaint grades,
And the event is set for each of the complaint grades.
The method according to claim 1,
Wherein the analysis unit obtains a consultation voice of an consultant who speaks with the customer and converts the complaint emotion into a numerical value by using a comparison result between a waveform of the customer voice and a waveform of the consultation voice.
The method according to claim 1,
Wherein the analyzing unit converts the complaint feeling into a numerical value by using a comparison result between a standard waveform of the call voice acquired from a plurality of customers and the customer voice.
The method according to claim 1,
Wherein the analysis unit digitizes the complaint emotion using a negative result of the negative speech included in the customer voice.
6. The method of claim 5,
Wherein the analyzing unit classifies the negative keywords into a plurality of negative classes to which different scores are assigned and quantifies the complaint feelings using the score of the class to which the negative word belongs.
The method according to claim 1,
Wherein the analysis unit includes a plurality of analysis means for grasping the complaint feeling and a collection means for collecting a plurality of output values output from the plurality of analysis means,
Wherein the combining means assigns different weights to the output values of the analyzing means and grasps the complaints of the customer comprehensively using the output values of the weighted analyzing means.
The method according to claim 1,
Wherein the analyzing unit includes a first analyzing unit, a second analyzing unit, a third analyzing unit, and a collecting unit,
Wherein the first analyzing means outputs a first output value obtained by digitizing a comparison result between a consultation voice of a consultant who speaks with the customer and the customer voice,
The second analyzing means outputs a second output value obtained by digitizing the comparison result between the standard waveform of the call voice acquired from the plurality of customers and the customer voice,
Wherein the third analyzing means outputs a third output value obtained by digitizing the negative speech included in the customer speech,
When the negative output signal is included in a first section in which the first output value satisfies the first set value or in the second section in which the second output value satisfies the second set value, Adding a first weight to the output value,
The combining means applies different second weights to the third output value to which the first weight is added, the first output value and the second output value,
And the collecting means collects the first output value, the second output value, and the third output value to which the second weight is applied, thereby to quantify the complaint feeling.
The method according to claim 1,
Wherein the classifying unit classifies the event corresponding to the customer's response plan into a plurality of steps,
Wherein the classifying unit sets a plurality of the complaint grades so as to match the event classified into a plurality of stages.
The method according to claim 1,
Wherein the processing section executes a complaint resolution event that lowers the customer's complaint emotion from the complaint level.
11. The method of claim 10,
Wherein the complaint resolution event displays a warning message on a consultation terminal of an agent who talks to the customer when the complaint grade meets a set grade.
11. The method of claim 10,
Wherein the complaint resolution event changes an agent who is in communication with the customer if the complaint rating satisfies a setting grade.
11. The method of claim 10,
Wherein the processing unit monitors the change of the complaint level according to the execution of the complaint resolution event and stores the monitoring result.
14. The method of claim 13,
Wherein the processing unit generates a correction message for the complaint resolution event if the complaint class does not fall despite the execution of the complaint resolution event.
The method according to claim 1,
Wherein the processing unit selects a new consultant to consult with the customer according to the degree of complaint and executes a connection event that connects the selected consulting terminal of the consultant to the customer terminal of the customer.
The method according to claim 1,
Wherein the processing section executes a consultation evaluation event for evaluating a consultant who is in consultation with the customer using the complaint grade.
The method according to claim 1,
And the processing section executes a customer evaluation event for evaluating the customer using the complaint level.

Images